of MoZecu~~r CQtu~ysis, 27 (1984)
213 - 224
CLAYS BY NEUTRON
T. J. PINNAVAIA Department
V. RAINEY, Materials
TZOU A ERE,
and J. W. WHITE* Physical
Summary Small-angle neutron scattering and inelastic neutron scattering have proved powerful methods for studying clay minerals - in particular platelet flocculation, particle and pore-size distributions and the dynamics of water and other included molecules. The usefulness of these methods for pillared clays is assessed here, showing how the pore and pillar sizes may be studied as well as the distribution function of the pillars. The likely dynamics of water in these systems is analysed with reference to model clay minerals and oxide catalyst supports.
Introduction The insertion of molecules into clay minerals to prop apart the aluminosilicate sheets, thereby producing larger pores than in native clays, was first done by Barrer and MacLeod [l]. Pillars of hydroxy aluminium and other cations capable of being dehydrated to oxide pillars and thereafter supporting temperatures of ca. 500 “C without structural collapse under catalytic cracking conditions are new, being first reported by Brindley and coworkers [2,3 ] and independently by Lahav et al. f4J and Vaughan and Lussier [ 51. Such materials work as catalytic crackers with an interesting pore-size distribution peaking in the 40 - 60 A region as well as below 14 A . They are relatively easily produced from fine particle-size clays such as bentonites or hectorites in suspension by treatment with partially hydrolysed salts such as A1C13~6H,0 in the form of ‘chlorhydrol’ always maintaining the OH/A ratio below 2.5 in the presence of Cl- or N03- ions at temperatures above 60 “C. There is considerable variation in final product texture, mechanical resistance and pore-size distribution; the origin of this and the mechanism of *Author
0 Elsevier Sequoia/Printed
in The Netherlands
platelet coalescence are susceptible to study by neutron small-angle scattering by analogy with simple colloids [6, 71 and other clays . We divide this preliminary study of these materials into four particular problems: (i) the aggregation of clay platelets to the polyhydroxy cation species; (ii) characterisation of the pillar and void sizes for polyhydroxy and dehydrated species; (iii) the pillar distribution function; and (iv) a discussion of the nature of water in the interlayer spaces by comparison with model systems.
and clay platelet
In Fig. l(a) the stages in the formation of a pillared clay, PILC, (from Vaughan and Lussier) are illustrated schematically. Despite the fact that the layer charge in smectite clays is highly irregular , the population density of polyhydroxy metal pillars is apparently the same in all layers [ 21, being determined by the radius of the hydrated cations whose charge, determined by the degree of hydrolysis, may in turn be prescribed by the layer charge [lOI* CLAY LAYER
CLAY LAYER ZAYLA$
? CLAY LAYER
i!gJiJ CLAY LAYER
[AL,304(OH)24(Hz0),~17+ 7H+ + 6tALz03+
8; Hz0 I
Fig. 1. Schematic diagram of the formation of dehydrated aluminium oxide pillared clay from sodium/calcium bentonite.
One mechanism for the process, illustrated in step I of Fig. 1, is that individual clay platelets in dispersion aggregate with the polyhydroxy cations. That this process can be studied uniquely well with neutron smallangle scattering is shown by measurements of the sensitivity of aggregation to cationic size . Clay spur bentonite (API No. 26), prepared by standard methods [ll], was converted to the homoionic forms containing Li+, Na+, K+ and Cs+ and the low-angle scattering patterns from 1% w/w aqueous solutions measured with 8 A and 4 A neutrons on the Dll instrument at Institut Laue-Langevin, Grenoble. Background and detector efficiency corrections were always made. Because the concentration was low, the scattering was from single particles. Since montmorillonite platelets are typically about 10 A thick and up to 2000 A in diameter, it is possible by choosing the range of momentum transfer of a neutron small-angle scattering experiment, to study selectively adsorption at the surface of the platelets and aggregation phenomena. For
is given by
r(Q) a R2Q2 exp(-Q2H2/12)/Q2
where R is the radius of the disc, N its thickness transfer in the scattering event.
and Q is the momentum
4lr Q = sin 6’ A where 0 is the half-angle of scattering. The formula is valid for QH < 1 and for QR > 1. The polydispersity in platelet diameters is not important in the range of Q used for these experiments. Figure 2 shows the pattern from a 1% w/w lithium montmorillonite solution. The linear portion of the curve in Fig. 2 is well represented by eqn. (l), where jF;T has a value of 10.3 8. This is the value expected for the completely dispersed sol. In contrast, as the counter ion is changed through sodium and potassium to caesium, there is a marked change in the small-angle scattering pattern and, in the case of a caesium sol with the same concentration of montmorillonite, the small-angle scattering is quite different (Fig. 3).
(Momentum transfer)‘, Q”, A-*
Fig. 2. Small-angle neutron scattering pattern from 1% w/w lithium montmorillonite water corrected for background and detector efficiency.
Fig. 3. Small-angle scattering from a 1% w/w solution of caesium montmorillonite in water corrected for background and detector efficiency. The slope of the log-log plot is -4 indicating that aggregation of the particles has occurred.
When H becomes large, the equation given above is no longer valid (CCtcit). Instead, the scattering is often given simply by Porod’s law 
216 As shown in Fig. 3, the scattering from the caesium sol obeys Porod’s law very well showing that H = 40 A and that extensive aggregation has occurred with particles containing, on average, about four platelets together. Potassium montmorillonite sols show similar scattering patterns to those of caesium. The scattering from sodium is close to that of lithium at high Q but shows some low Q effects. Insofar as it has been possible to detect clustering of platelets of caesium montmorillonite, it is evident that the study of adsorption on the platelets is readily accessible through lowangle neutron scattering; the platelet surface area is of the order of 800 m* mineral g-l and the contrast matching point [lo] for the montmorillonite lies between 60 and 70% D,O, depending on whether a full proton exchange or no proton exchange with the protons in the mineral occurs. We suppose that it would be of interest to make a study similar to the above on solutions of hydroxy aluminium salts and clay suspensions at different stages in their maturation, to distinguish this mechanism from others involving direct substitution of simple cations.
from pillars and voids in pillared clays
Neutron scattering arises from discontinuities in the scattering length density of a material - in the above example, from the individual platelets whose neutron coherent scattering length per unit volume, ps, is different from the surrounding water. Table 1 gives the neutron scattering length densities for species relevant to the study of pillared clays. From a dry pillared clay powder, there will be discontinuities in density of various scales; firstly, due to the grain size of the whole clay particle; secondly, as between the clay mineral and the pores within it, and thirdly, between the pillars, their TABLE
2.8 ‘2.0’ 2.0 3.5 1.0 1.1 0.88 0.88
‘A~~~~~(OD)Z~(DZO)IZ: A~I~~~(OH)~~(HZO)I~ 4203 H2O D2 C6H6 C6D6
“ps = i
pS, of clay materials
22.65 59.77 17.3 2.44 -0.17 1.91 1.75 8.0
2.70 6.6 1.3 4.8 - -0.56 6.36 1.2 4.8
surrounding pores and the adjacent clay mineral. By reference to Table 1, we can see that the contributions from these separate distributions may be selectively suppressed by contrast matching. For example, if we fill the spaces between clay particles with a mixture of C6H, and C6D6, or even Hz and D2, for which pS equals that of the clay, there will be no scattering length density discontinuity and intercrystalline scattering is suppressed. Equally, scattering from Al,O, pillars can be completely suppressed by filling the intracrystalline regions with C,D,. In the experiments below, pure benzene alone was used since the degree of deuteration of the aluminium polymer species was not certain - thereby reducing the clay particle and intracrystalline void scattering relative to that from ‘Al,304(OD)24(D20)12’ or Al*O, pillars depending upon the sample.
Experimental The pillared clays were prepared by the reaction of Na+-exchanged montmorillonite with chlorhydrol. One sample was air-dried, exchanged with D20 vapour and then freeze-dried. This sample was never heated beyond 70 “C and should still be hydroxylated. The other sample was heated to 350 “C and should be totally dehydroxylated. The samples containing benzene were made up by wetting the powder thoroughly with dry AnalaR reagent in the 1 mm thick silica sample cells. Small-angle neutron scattering experiments with 8 A and 5 A neutrons were made at the PLUTO reactor. AERE, Harwell, the sample being sealed from the surrounding vacuum and kept at 25 “C. For the clays without benzene, both samples were 0.6 cm thick and 8 A neutrons gave a Q range of 0.01 8-i < Q < 0.8 8-l. For these experiments the transmissions were 0.35 (anhydrous sample) and 0.25 (deuterated sample), which confirmed that the latter had incomplete hydrogen-deuterium exchange. formula of the pillard montmorillonite is The unit cell
Feo.42Wh.481 [Si7.ssAl,.121020(OH)4.The BET surface [Al(OH)2.8o12.87[A13.~~ area after activation
at 360 “C was 260 m2 g-i.
R.esults Without contrast variation The dependence on Q of the intensity of small-angle scattering from roughly spherical density discontinuities  is best characterised by the is mean-square radius of gyration, Rg2, which in the Guinier approximation related to the scattering intensity at a particular Q value by I(Q) = const exp(-R,‘Q’)
For a delta function distribution of sizes, a straight line results from a plot of In I versus Q2. It can be seen that this is not so for either of the dry clay
-l-oool (a) ANHYDROUS
‘.., .. .. . . . ..
1.0 lo2 x (Momentum
‘.. . . . , . .
Fig. 4. Small-angle scattering pattern from (a) dehydroxylated A1rs7+ pillar montmorillonite (8 A incident neutrons) and (b) the pattern from DzO-exchanged and freeze-dried 7c montmorillonite (8 A incident neutrons). Ml3 Fig. 5. Difference-pattern between small-angle patterns from DzO-exchanged and dehydroxylated A1ra7+pillar montmorillonite.
samples. Figures 4(a) and 4(b) show Guinier plots (In I uersus Q2) for the anhydrous and D,O clay respectively. Figure 5 shows the difference pattern for D20-exchanged clay-anhydrous clay, a difference which should suppress somewhat the clay crystallite scattering. The D20-exchanged clay has the more intense scattering pattern, but for both there is clearly a range of particle sizes giving the scattering. Fitting straight lines to the initial slopes 0 < Q2 < 1 X low3 Aw2, and final slopes, 0.8 X 10P2 < Q2 < 1.6 X lop2 A--‘, gives extrema for the Guinier radii for the two materials as shown in Table 2. We can conclude that to within the errors of the present measurement the TABLE 2 Guinier radii for anhydrous DzO-exchanged pillar clay and for the difference (DzO-exchanged clay-anhydrous clay) Momentum transfer value Q = 0.02 initial slope Q= 0.1 final slope
DzO-exchanged clay-anhydrous clay
119.8 + 3.4
115.2 + 3.3
16.6 f 0.3
16.1 f 0.2
16.6 f 0.2
Guinier radii are the same for the anhydrous and DzO-exchanged clays. In the Q range studied, there does not appear to be a strong contribution to scattering from intercrystalline voids in the powder packing. With con trust variation To further clarify the origin of the strong small-angle scattering in Figs. 4 and 5, the pores of both clay samples were filled with varying amounts of benzene. For the cavities filled with C6H6, the effect expected from Table 1 is to halve the contrast between clay mineral and hitherto empty intracrystalline pores (p, = 0), thereby emphasising scattering from the shape of the pillars - and to some extent the scattering from the pillarpillar distribution function (see below). The samples were also made 0.1 cm thick so as to minimise multiple scattering. Figure 6 shows the Guinier plot for 8 a incident neutrons on a 1 mm thick sample of the deuterium-exchanged clay sample immersed in liquid C,H6. The curve for the anhydrous sample is very similar in appearance, though with weaker scattering. The curves have been fitted in the Guinier approximation to give radii of gyration at low and high Q of ea. 113 ,& and 12.5 a. The latter corresponds fairly well with the average pore radius determined by porosimeter  (11.8 8) although the Porod region is being approached. The relative effect of adding C,H, to the low-angle and highangle parts of the pattern was different, the low-angle intensity being somewhat more sensitive. This suggests that the full contrast effect was not seen, possibly due to occluded air in the narrower pores and that vacuum filling or smaller molecules are thus necessary. The results are summarised in Table 3, the larger radius being probably a measure of the average size of the scattering particles. The variation of the root-meansquare radius of gyration with decreasing contrast may be noted by comparing Table 2 and Table 3. For
, ---y-h_ 0880 10Z~(Momentum
Fig. 6. Guinier plot of the small-angle clay suspended in liquid benzene.
Guinier radii for anhydrous pillared clay and DzO-exchanged liquid benzene to achieve some contrast matching Momentum transfer value Q = 0.02 initial slope Q= 0.1 final slope
clay + liq. C6Hh
clay + liq. CGH~
113 * 3
the anhydrous clay, there is a drop in (R,) from both low Q and higher Q slopes, indicating a peaked form for the average scattering potential function for those species causing this scattering. But for the DzO-exchanged clay, although CR,) dropped at low Q the 16.6 A value without benzene showed a constancy or possible increase as the contrast was lowered. In view of the incomplete deuteration of the hydroxylated pillars this must be regarded as a provisional result, but it does show that the effective potential in this system is of a different shape. This suggests the important conclusion that systematic contrast variation studies may give details of the pillar shapes. Measurements in the diffraction region Because of inherent disorder and the small particle size of pillared clays, the diffraction pattern is broad. Nevertheless for polyhydroxy aluminium clays, for example, the scattering contrast between pillars and clay mineral is much greater for neutrons than for X-rays, especially when these pillars are decorated by exchange with deuterium. Figure 7 shows the scattering intensity (on a linear scale plotted uersus Q) for the 0.6 cm thick dehydroxylated clay sample taken with 5.0 A neutrons. The peak at Q = 0.365 A-* corresponds to a d-spacing of 17.21 a and is the (001) reflexion. Its breadth (cu. 0.1 A-‘) indicates a coherence length (or crystallite size) along the c-axis of about 120 _!%. In order to get the relative intensitives of small-angle scattering and the very weak peak into perspective, the results for both anhydrous clay and DzO-exchanged clay have been replotted with a logarithmic ordinate scale in Figs. 8(a) and (b), respectively. The presentation of Fig. 8 also makes it clear that in the DzOexchanged clay system there is not only a peak in the region of 18 A once again, but there is additional scattering filling up the valley to lower Q (0.1 8-l < Q < 0.3 a-‘). A study with variable contrast would allow this scattering to be extracted systematically, but since the patterns have been determined on an absolute scale this second diffraction component is here extracted approximately by subtraction of the anhydrous pattern from the deuterio polyhydroxy clay pattern. The procedure is limited in precision by
OOOoop 10x Momentumtransfer, Q. %-’ Fig. 7. Diffraction
pattern from anhydrous
AI pillared ciay taken with 5.0 a neutrons.
Fig. 8. Diffraction region and small-angle regions for (a) the dehydroxylated and (b) the [email protected]
Al13~+pillar montmorillonite. (The ordinate is a logarithmic scale.)
an effect due to the slight change in the c-axis lattice parameter, hut if we take the Bragg peak at d = 18 _&.to be about the same intensity in the two patterns subtraction gives the pattern shown in Fig. 9 (curve A). This curve shaws a peak at ea. Q = 0.27 a-l and possibly Q = 0.5 A-” (d-spacings 23.2 a and 12.6 8, respectively). In Fig. 9 we have also made an attempt to subtract out the tail of the small-angle scattering, by modelling it with a flat background and an empirical function I = F(Qw2). When this pattern (curve B) is subtracted from curve A we are left with a dist~ibnti~n (curve A-3). Discussion The peak at Q = 0.25 in the difference pattern between deuterated and anhydrous clay is most likely to have arisen from the lateral packing of
10 x Momentum transfer, Q, 1-l Fig. 9. (A) Difference between normalised diffraction patterns (background subtracted) (B) model of the for DzO-exchanged and dehydroxy~ated A.l,37+ pillar montmoriilonite. between model and measured small-angle scattering component; (A-B) difference scattering.
pillars in the deuterated material (i.e. before ignition). The great breadth of the peak, AQ = 0.15 a-‘, makes it more reminiscent of the first peak in a liquid or amorphous structure factor. On this hypothesis the average spacing between pillars would be ca. 25 a. Alternatively, if it is the (020) reflexion or if the pillar packing is approximately hexagonal, the mean interpillar distances would be ca. 50 a and cu. 29 8, respectively. Treated as a Bragg peak, the width indicates a lateral coherence length for the pillar packing of about 170 8. Further work at higher neutron intensities is needed to study the relative intensities of this peak and the possible weak feature at Q = 0.5 A-‘. A preliminary estimate of pillar lateral radius is ea. 4 A from the structure factor. Ordering
of the water in pillared
The extent to which water and dissolved ions are immobilised as a function of surface properties, e.g. surface charge density, counter ion charge, H-bond formation, and the thickness of the water layer above the surface, has recently been reviewed [ 141. These results may be used to form a preliminary insight into the state of water in pillared clays. Because of the presence of both alumina and silicate surfaces in these systems, the physical properties of the residual water, which seems to play a vital part in their catalytic behaviour, is likely as a first approximation to be a superposition of characteristics already found [ 15,161 for simple smectites and hydrophilic oxides [ 14,17 1. Using neutron quasi-elastic scattering, one may distinguish motions of a molecule or atom on the timescale of cu. lo- l”s. This is a particularly apt timescale to see first and second hydration shell water in solutions and clay minerals. For example, water molecufes which are held for times of ea. lo-’ - 10W9 s (first shell for polyvalent cations) are
seen as bound
waters by this method, but those which are more mobile give strong quasi-elastic energy broadening. Distinctions inaccessible to longer timescale methods, e.g. NMR spectroscopy, may be drawn. What are these characteristics on the neutron timescale? Apart from physical effects due to cage size, there seems to be very (a)
little bound water in monovalent ion-exchanged smectites, e.g. montmorillonite, vermiculite, after swelling by more than the equivalent of three monolayers of water. Modification of the diffusion coefficient is readily detectable even in clays swollen to an intracrystalline water layer thickness of cu. 50 8, but the diffusion constant values are only about 50% less than for pure water at corresponding temperatures. Furthermore the diffusion tensor shows little anisotropy. The conclusion is that extensive ice-like structures are not formed and that there is rapid exchange on the lo-’ s timescale between intracrystalline water molecules in different sites. At low hydration (one or two equivalent monolayers of water), translation diffusion and rotational diffusion are both seen for the hydration waters of the counter ions. These more restricted motions are alone seen for bivalent ion-exchange clays in the crystalline swelling region. In higher surface charge density layer systems, such as lyotropic liquid crystals where ionised carboxylate groups form the solid/liquid boundary, the neutron quasi-elastic scattering pattern is better described by the superposition of broad and narrow peaks corresponding to free and bound water. Even in these systems a clear separation of these peaks is not seen under conditions of 30 PeV energy resolution. This again indicates that exchange is fast. Once again the feeble anisotropy of the water diffusion tensor is seen (with greater reliability here because much more perfect layer orientation is produced than in clays). Also, again the relatively small structuring effect on the water diffusion caused by the charged interface is confirmed. In sols and gels of fumed silica, a qualitatively different picture emerges from the quasi-elastic scattering. Bound and free water peaks are immediately visible from the quasi-elastic spectrum even at 30% w/w H,O/SiO, ratios. Localisation of Hz0 by siloxyl groups is evidently responsible, and a much stronger structure-making phenomenon than that existing at clay/water or liquid crystal/water interfaces.
Conclusions Some new information about the ordering and shape of pillars in polyhydroxy aluminium and aluminium oxide pillared montmorillonite is accessible by neutron small-angle scattering, and the process of formation of these materials may also be open to more detailed study.
The dynamics of water in swollen smectites, including pillared clays, more closely resembles bulk water than ice - from the point of view of its mobility - except at certain highly charged sites and at hydrations corresponding to up to a few statistical monolayers coverage.
References 1 R. M. Barrer and D. M. MacLeod, Trans. Faraday Sot., 51 (1955) 1290. 2 G. W. Brindley and R. E. Sempels, Clay Miner., 12 (1977) 229. 3 S. Yamanaka and G. W. Brindley, Clays Clay Miner., 26 (1978) 21. 4 H. Lahav, LJ.Shani and J. Shabtai, Clays Clay Miner., 26 (1978) 107. 5 D. E. W. Vaughan and R. J. Lussier, Prepr. 5th Znt. Conf. Zeolites, Naples, 1980. 6 D. J. Cebuia, R. K. Thomas, N. M. Harris, J. Tabony and J. W. White, Discuss. Faraday Sot., 65 (1978) 76. 7 N. M. Harris, R. H. Ottewill and J. W. White, in R. H. Ottewill, C. Rochester and A. L. Smith (eds.), Adsorption from Solution, Academic Press, London, 1983, p. 139. 8 D. J. Cebula, R. K. Thomas and J. W. White, Deu. Sedimentol., 27 (1979) 111. 9 See, for example: G. Lazaly and A. Weiss, in S. W. Bailey (ed.), Prepr. Znt. Clay Conf., Mexico City, 1975, Applied Publishing Ltd., Wilmette, IL, 1976, pp. 157 - 172. 10 T. J. Pinnavaia, Science, 220 (1983) 365. 11 I. C. Callaghan and R. H. Ottewiil, Discuss. Faraday Sot., 57 (1974) 110. 12 A. Guinier and G. Fournet, Small-angle Scattering of X-rays, Wiley, New York, 1955. 13 B. E. Warren,Phys. Reu., 59 (1941) 693. 14 J. W. White, in Vu Thien Binh (ed.), Surface Mobilities on Solid Materials, Plenum, New York, 1983, p. 527. 15 D. J. Cebuia, R. K. Thomas and J. W. White, Clays Clay Miner., 29 (1981) 241. 16 P. L. Hall, D. K. Ross, J. J. Tuck and M. H. B. Hayes, Neutron Inelastic Scattering, International Atomic Energy Agency, Vienna, 1977, pp. 617 - 633. 17 J. Ramsay and R. B. Richardson, Reported at the Workshop on the Structure of Water at Interfaces, Institut Laue-Langevin, Grenoble, 1981.